The jaguar (Panthera onca) and puma (Puma concolor) are the largest felids of the American Continent and live in sympatry along most of their distribution. Their tracks are frequently used for research and management purposes, but tracks are difficult to distinguish from each other and can be confused with those of big canids. We used tracks from pumas, jaguars, large dogs, and maned wolves (Chrysocyon brachyurus) to evaluate traditional qualitative and quantitative identification methods and to elaborate multivariate methods to differentiate big canids versus big felids and puma versus jaguar tracks (n  =  167 tracks from 18 zoos). We tested accuracy of qualitative classification through an identification exercise with field-experienced volunteers. Qualitative methods were useful but there was high variability in accuracy of track identification. Most of the traditional quantitative methods showed an elevated percentage of misclassified tracks (≥20%). We used stepwise discriminant function analysis to develop 3 discriminant models: 1 for big canid versus big felid track identification and 2 alternative models for jaguar versus puma track differentiation using 1) best discriminant variables, and 2) size-independent variables. These models had high classification performance, with <10% of error in the validation procedures. We used simpler discriminant models in the elaboration of identification keys to facilitate track classification process. We developed an accurate method for track identification, capable of distinguishing between big felids (puma and jaguar) and large canids (dog and maned wolf) tracks and between jaguar and puma tracks. Application of our method will allow a more reliable use of tracks in puma and jaguar research and it will help managers using tracks as indicators of these felids' presence for conservation or management purposes.